JP2001522174A - Digital image processing method and processing apparatus - Google Patents

Abstract

(57) [Summary] MPEG Standard, H. 261 or H.264. In transmitting a motion image according to the H.263 standard, an area that satisfies a predetermined criterion of an image to be transmitted is encoded with higher image quality than other parts of the image. This region meets the criteria when it has a predetermined color, preferably a color similar to human skin. High image quality in regions that meet the criteria is ensured by relatively small quantization values, relatively high positional resolution, or relatively high image repetition rates.

Description

DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to a digital image processing method and a digital image processing apparatus.

[0002] In particular, when digitally transmitting an image by conventional image compression, that is, reduction of image data to be transmitted, a channel for transmission becomes a bottleneck. The image data is transmitted via this channel from the (image) coder to the (image) decoder.
The bandwidth of this channel is usually predetermined and constant. The compression is accordingly adjusted in the coder to this bandwidth. Standard block-based image coding methods such as MPEG-4 or H.264. According to the means provided by H.263 (see [1]), the coder can be pre-tuned or adapted so that the number of images transmitted per time unit is guaranteed. Here, a decrease in image quality is tolerated.

[0003] One example is an image phone that displays images. This image is transmitted over the ISDN B channel at a transmission rate of 64 kbit / s and has a corresponding image quality. The image is always displayed in small, relatively small pieces in a low-resolution image sequence.

[0004] Block-based image coding method (eg, MPEG standard or H.263 standard)
Is known from [2].

In image processing, image quality is set for the entire image to be transmitted. Depending on the data rate used, the image quality is adapted so that the bandwidth is used efficiently, but the entire image can be transmitted. The disadvantage here is that in order to transmit the entire image, a reduction in image quality must be tolerated.

An object of the present invention is to provide an image processing method and an image processing apparatus which avoid the above-mentioned disadvantages.

This problem is solved by the configuration according to claims 1 and 20.

In the image processing method of the present invention, the digital image is divided into two regions, where the first region satisfies a predetermined criterion and the second region does not satisfy the predetermined criterion. The first area is then processed with a relatively high image quality.

For the first region, an improvement in image quality is achieved, since this satisfies a predetermined criterion. Thus, according to the invention, the first region can be electronically processed with high quality as part of the whole image, for example transmitted or compressed.

In an improved embodiment of the present invention, the image is divided into a plurality of regions and a plurality of first
Are processed with higher image quality than a plurality of second regions that do not satisfy the criterion.

Thus, the plurality of regions in the image are divided into a plurality of first regions and a plurality of second regions, so that even if the first regions in the image are not connected, a relatively high image quality can be obtained. Can be assigned.

In another refinement, the criterion is met when the first region of the image has a predetermined color. This can be, for example, a color close to the color of human skin.

In an additional refinement, the image is processed by a block-based image coding method. Examples for block-based image coding methods are the MPEG standard or H.264. 2
An image coding method defined by the H.63 standard.

In the block-based image coding method, for each block of the image, which block has a predetermined size, the color of the block is in the form of a second color value, advantageously the average of the pixels of this block Is detected by A comparison operation is performed between the second color value and the first color value. If the result of the comparison operation is smaller than a predetermined threshold, the criterion is fulfilled, so that the block has at least a color similar to human skin. Otherwise (
If the result of the comparison operation is not less than a predetermined threshold), the criterion for this block is not met.

In an additional refinement, the method is performed iteratively on each block of the image.

The predetermined size of the block is advantageously 8 × 8 or 16 × 16 pixels.

The comparison operation can be defined in various forms. The following three means (formula (1)
To (3)).

| Xy-hy | + | xCr-hCr | + | xCb-hCb | <S (1) | xy-hy | ² + | xCr-hCr | ² + | xCb-hCb | ² <S (2) k1 | D1 | + k2 | D2 | + k3 | D3 | <S (3) where xy is the luminance value (= lightness) of the first color value xCr the first chrominance value of the first color value (= hue) XCb Second chrominance value of first color value (= saturation) hy Luminance value of second color value (= brightness) hCr First chrominance value of second color value (= hue) hCb second The second chrominance value (= saturation degree) of the color value of S is the predetermined threshold value k1, k2, k3 The predetermined weight value D1 The first comparison between the luminance value of the first color value and the luminance value of the second color value D2 a second comparison of the first chrominance value of the first color value with a first chrominance value of the second color value D3 a second comparison of the second chrominance value of the first color value and the second chrominance value of the second color value 3 represents a third comparison with a chrominance value of 2.

It is particularly advantageous if the three parameters (luminance value, first chrominance value, second chrominance value) are determined pixel by pixel or image block (appropriately averaged from the pixels of the image block) at a time. Is compared with the threshold value. Similarly, it is conceivable to compare the brightness, hue and saturation for each block of the image with predetermined values belonging to the color of the skin. In this case, the results of three comparison operations are obtained, and these three results can be used to determine whether the criteria are met. In other words, individual comparison results are obtained for D1, D2, and D3. From the above equation (1),
In 3), the three individual comparison values are combined with one another and compared as a whole with a predetermined threshold value.

Further, a block-based image coding method according to MPEG-4 can be used. Here, according to the MPEG-4 standard, the transmission format can be unified for a predetermined area (so-called image object), that is, for an area where a predetermined standard is satisfied.

In another refinement, a relatively high image quality is achieved as follows. That is, a relatively high image repetition rate is determined for the first region. By selecting a higher image repetition rate for the first area (or correspondingly more first areas in the digital image) than for the (at least one) second area, the first Area is updated particularly frequently, so that the motion is displayed smoothly. When displaying a face like a newscaster, the observer sees the movement of the lips in a continuous motion image, whereas the background that does not correspond to the color of the skin is hardly updated, and therefore the motion in the background Is perceived in pieces.

Another means for improving the image quality for the first area is to increase the position resolution for this first area. In this way, a relatively large number of pixels for the second area ensures a clear display of the first area.

A third means for improving image quality is to control the quantization value for both the first region and the second region. Here, the first area is quantized by the first quantization value, and the second area is quantized by the second quantization value. Relatively high image quality of the first region is ensured by making the first quantization value smaller than the second quantization value.

In the framework of the additional refinement, according to the MPEG standard, a first quantization value (relatively small value → high resolution → high demand for bandwidth) is replaced by a second quantization value (relatively high value). (Larger value → lower resolution → lower demand for bandwidth) (here, the first quantization value is smaller than the second quantization value). This switching is done by DQUAN
This is performed by transmitting the T symbol as a parameter together with the second quantization value.
Conversely, switching from the second quantized value to the first quantized value is performed by transmitting a DQUANT symbol as a parameter together with the first quantized value.

In a refinement of the invention, the relevant area as large as possible is guaranteed as the first or second area. This is because, in this way, the process of switching between quantization values can be reduced. As described above, DQUANT is added to each switching process.
The symbol precedes. This symbol forms an overhead, which results in a loss of bandwidth that could otherwise be used for image information.

If a plurality of blocks of the image satisfying the criterion are processed with corresponding quantized values, they are between these blocks and exceed a predetermined threshold value according to one of equations (1) to (3). Individual blocks must be quantized as image blocks that meet the criteria with different quantization values. It is therefore advantageous to treat individual blocks, which have to be combined with special new quantization values, with relatively small quantization values as if they meet the criteria.

The aim is to reduce the switching between the quantized values which respectively apply to the subsequent blocks of the image. Individual blocks in the blocks that meet the criteria are therefore advantageously processed with relatively small quantization values and transmitted with correspondingly high image quality, as are the surrounding image blocks.

In the pre-processing prior to the original quantization, which blocks meet the criterion, which blocks are near the blocks that meet the criterion, and which are near the blocks that do not meet the criterion, It is advantageous to set whether a predetermined interval is set for blocks that do not satisfy the condition and satisfy the criterion. In this way, an advantageous number and distribution of the switching processes between the quantized values can be determined on the basis of a block-based coding method corresponding to the line-by-line processing of the image. This is done by detecting areas that are as relevant as possible to blocks of the image that meet or do not meet the criteria.

For this purpose, the threshold value S for the block is controlled as a function of the number that exceeds this threshold value S in the neighboring blocks. The threshold value S is advantageously reduced when a large number of neighboring blocks have a given color and correspondingly increased when a small number of neighboring blocks have a given color.

It is also possible to apply the image quality improving means in combination. Most of the desire to improve image quality is limited, for example, only to conditions within the physical frame of maximum bandwidth used. This depends only on the respective application and does not limit the combination of image quality improvement measures, in particular the improvement of the area of the digital image.

The gist of the present invention is to set the image quality locally differently in the transmitted image. This is very advantageous because image parts close to the skin color can be transmitted with higher image quality than other parts of the image. An example of this is a news broadcast, in which the newscaster's face and hands are quantized and transmitted with a relatively smaller quantization value than non-skin-colored image areas, ie with high image quality.
Therefore, even for the hearing impaired, information can be obtained from high-quality image areas (skin color areas).
For example, it can be obtained by lip reading or gesture interpretation. In particular, this makes visual telephony an effective communication medium for speech or hearing impaired persons. This is because hands and facial expressions (lips) are transmitted with a higher quality than the rest of the image, thus providing a means for even speech or hearing impaired people to communicate by lip movements or hand movements.

The present invention further provides an image processing device using a processor unit. The apparatus is configured such that the method steps described above are performed.

Further, an apparatus for encoding a digital image is provided. The device comprises means for performing a spectral transformation, which converts the digital image into the spectral domain. Means for entropy coding can also be provided, which perform data compression. In addition, the coding device has a buffer, which receives data of variable data rate from the means for entropy coding and preferably further communicates with the fixed data rate channel.

In a further refinement of this initial diameter, the filling state of the buffer is adapted to the quantizer quantization value.

In another refinement, the adaptation of the quantizer is performed as follows. That is,
A filled buffer increases the quantization value, causing the image to be blurred, and an empty buffer reduces the quantization value, ensuring high image quality.

[0036] Improvements of the invention are evident from the dependent claims.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a device having two calculators, a camera and a screen, by which the encoding, transmission, decoding and display of the image data takes place.

FIG. 2 is a schematic diagram of an apparatus for encoding a digital image on a block basis.

FIG. 3 shows the steps of a method for quantizing a digital image by a calculator.

FIG. 4 is a schematic diagram showing how quantization is adapted depending on the buffer state.

FIG. 5 is a schematic diagram of a means for improving image quality.

FIG. 1 shows an apparatus having two calculators and a camera. Here, image encoding, transmission of image data, and image decoding will be described.

The camera 101 is connected to the first calculator 102 via a line 119. The camera 101 transmits the recorded image 104 to the first calculator 102. The first calculator 102 has a first processor unit 103, which is connected to the image memory 105 via a bus 118. First calculator 1
The image coding method is performed by the processor unit 103 of the second embodiment. The image data 106 encoded in this format is transmitted from the first computer 102 to the communication connection path 107,
Preferably, it is transmitted to the second calculator 108 via a line or a wireless path. The second calculator 108 includes a second processor unit 109, which is connected to an image memory 111 via a bus 110. In the second processor unit 109, an image decoding method is performed.

Both the first calculator 102 and the second calculator 108 have screens 112 to 113, respectively.
, On which the image data 104 is visualized. An input unit is provided for operating the first calculator 102 and the second calculator 108, respectively, which are preferably keyboards 114 to 115 and computer mice 116 to 117.

The image data 104 transmitted from the camera 101 to the first calculator 102 via the line 119 is preferably data in the time domain. On the other hand, the first calculator 102
The data 106 transmitted via the communication connection 107 to the second calculator 108 is preferably image data in the spectral domain.

The screen 120 displays the decoded image data.

The following briefly describes MPEG image coding. For details, refer to the description of [2].

The encoding method used in the MPEG standard is substantially based on the hybrid DCT (Discrete Cosine Transformation) with motion compensation. This method is used for n × 64 kbit / s video telephony (CCITT Recommendation H.261) and for 34 to 45 Mbit / s TV contribution (
CCR Recommendation 723) and for 1.2 Mbit / s multimedia applications (ISO
-MPEG-1) used in a similar form. The hybrid DCT is composed of a temporal processing stage using similarity between successive images and a positional processing stage using correlation in the image.

Positional processing (intra-frame coding) substantially corresponds to classical DCT coding.
The image is decomposed into 8 × 8 pixel blocks, and each of these blocks is transformed into the frequency domain by DCT. The result is an 8 × 8 coefficient matrix, which approximately reflects the two-dimensional spatial frequency in the transformed image block. The frequency of the AC coefficients rises along the fx and fy axes, while the DC coefficients (DC components) represent the average gray value of the image block.

After the conversion, data extension is performed. This is because the coefficient amplitude is advantageously sampled at 12 bits for reasons of computational accuracy. Of course, in a natural original image, the energy is concentrated around the DC value (DC component), and the coefficient of the highest frequency is usually zero.

In the next step, spectral weighting of the coefficients is performed. As a result, the amplitude accuracy of the high frequency coefficient decreases. Here, the characteristics of human eyes are used. The human eye has lower resolution at higher spatial frequencies than at lower spatial frequencies.

The second step of data compression is performed in the form of adaptive quantization. This adaptive quantization further reduces the amplitude accuracy of the coefficients, whereby small amplitudes are replaced by zero. The degree of quantization here depends on the filling state of the output buffer. If the buffer is empty, fine quantization is performed, thereby forming more data. On the other hand, if the buffer is full, coarse quantization is performed, thereby reducing the amount of data.

After quantization, the blocks are sampled diagonally (zigzag scanning), followed by entropy coding. The original data compression is performed by this encoding. Two effects are used for this.

1) Statistics of amplitude values. Higher amplitude values are less frequent than lower amplitude values, so rare events are assigned long code words and frequent events are assigned short code words (
Variable length coding, VLC). Thus, on average, the data rate is lower when encoding with a fixed word length. The variable speed of the VLC is subsequently smoothed in the buffer memory.

2) Take advantage of the fact that often only zeros follow from a given value. Instead of all zeros, simply transmit an EOB code (End of Block). This makes the encoding gain significant when compressing image data. Departure speed 51
Instead of 2 bits, in the above example, 46 bits are transmitted for this block.
This corresponds to a compression factor of 11 or more.

Further compression rates are obtained by temporal processing (intra-frame coding). The encoding of the difference image requires a lower data rate than the encoding of the original image. This is because the amplitude value is extremely small.

When the motion in the image is small, the time difference is small. However, when the movement in the image is large, the difference that occurs is large, and it is difficult to encode the difference. For this reason, the motion for each image is measured (motion prediction) and compensated before the difference formation (motion compensation). Here, the motion information is transmitted together with the image information. Usually, a motion vector per macroblock (for example, four 8 × 8 image blocks) is used.

The use of motion compensated bidirectional prediction instead of the prediction used results in even smaller amplitude values of the difference image.

In the motion-compensated hybrid coder, the temporal difference signal is converted instead of the image signal itself. For this reason, the coder uses a temporal repetition loop. This is because the predictor has to calculate the predicted value from the values of the already transmitted (encoded) image. The same temporal repetition loop exists in the decoder, and the coder and decoder are completely synchronized.

There are mainly three different MPEG2 coding methods that can perform image processing.

I-picture: No temporal prediction is used in the I-picture. That is, the image values are directly transformed and encoded. This is shown in image 1. The use of I-images is
This is to enable a new start of the decoding process without prior knowledge or to achieve resynchronization in case of transmission errors.

P-picture: Temporal prediction is performed based on the P-picture. DCT is applied to the temporal prediction error.

B-picture: For B-pictures, the prediction error in both directions is calculated in time and subsequently transformed. Bidirectional prediction basically works adaptively. That is, forward prediction, backward prediction, or interpolation is allowed.

When the image sequence is MPEG-2 encoded, a so-called GOP (Group Of Pic
tures). The n pictures between two I-pictures form one GOP. The interval between P-images is represented by m. Here, there are m-1 B-pictures between the P-pictures. MPEG syntax leaves the choice of m and n to the user. m = 1 means that no B-picture is used, and n = 1 means that only the I-picture is coded.

FIG. 1 is a schematic diagram of an apparatus for performing a block-based image coding method according to the H.263 standard (see [1]).

A video data stream to be coded comprising digital images that are successive in time is supplied to the image coding unit 201. The digital image is divided into macro blocks 202. Here, each macro block has 16 × 16 pixels. The macro block 202 includes four image blocks 203, 204, 205, and 206.
Here, each image block includes 8 × 8 pixels, and a luminance value (brightness value) is assigned to each pixel. Further, each macroblock 202 has two chrominance blocks 2
07 and 208, these blocks have chrominance values (color information, saturation) assigned to the pixels.

A block of one image includes a luminance value (= brightness) and a first chrominance value (= hue)
And a second chrominance value (= saturation). Here, the luminance value, the first chrominance value, and the second chrominance value are referred to as color values.

The image block is supplied to the transform coding unit 209. In the case of differential image coding, the value to be coded of the image block of the temporally preceding image is subtracted from the value of the instantaneous image block to be coded. Then, only the difference forming information 210 is supplied to the transform coding unit (discrete cosine transform, DCT) 209. For this purpose, the instantaneous macroblock 202 is connected via the connection 234 to the motion prediction unit 229.
Will be notified. In the transform coding unit 209, spectral coefficients 211 are formed for the image block to be coded or for the difference image block, and are supplied to the quantization unit 212. This quantization unit 212 corresponds to the quantization device of the present invention.

The quantized spectral coefficients 213 are supplied to both the scan unit 214 and the inverse quantization unit 215 in the feedback path. After scanning, for example by a “zigzag” scanning method, entropy coding is performed on the scanned spectral coefficients 232 in an entropy coding unit 216 provided for this. The entropy-coded spectral coefficients are transmitted as coded image data 217 to the decoder via a channel, preferably a line or a radio link.

The inverse quantization unit 215 performs inverse quantization of the quantized spectral coefficients 213. The spectral coefficient 218 thus obtained is converted to the inverse transform coding unit 2
19 (inverse discrete cosine transform, IDCT). The reconstructed encoded value (and also the differential encoded value) is supplied to the adder 221 in the differential image mode. The adder 221 further receives the encoded value of the image block. This coding value is obtained from the temporally preceding picture according to the motion compensation already performed. An image block 222 restored by the adder 221 is formed and stored in the image memory 223.

The restored chrominance value 224 of the image block 222 is supplied from the image memory 223 to the motion compensation unit 225. Interpolation is performed on the brightness value 226 by an interpolation unit 227 provided for that purpose. Based on the interpolation, the number of lightness values contained in each image block is doubled. All brightness values 228 are provided to both the motion compensation unit 225 and the motion prediction unit 229. The motion prediction unit 229 also receives via a connection 234 an image block of a macroblock (16 × 16 pixels) to be coded, respectively. In the motion prediction unit 229,
Motion prediction is performed in consideration of the interpolated brightness value (half-pixel based motion prediction). Advantageously, in the motion estimation, the absolute difference between the individual brightness values in the instantaneous macroblock 202 to be coded and in the macroblock recovered from the temporally preceding picture is detected.

The result of the motion prediction is a motion vector. With this motion vector, the macroblock selected from the temporally preceding image and the macroblock 20 to be encoded
2 is represented.

Both the brightness information and the chrominance information are shifted by the motion vector 230 with respect to the macro block obtained by the motion prediction unit 229, and the macro block 20
2 (see data arrow 231).

FIG. 3 shows method steps for quantizing a digital image with a calculator.

In step 301, flesh color COL _H is detected as a predetermined color (reference for the first area of the image). In step 302, a block of an image is read and its color COL _BL is detected. Since each individual pixel of the block has a lightness value, a hue and a saturation, a corresponding average value is formed for the whole block. This average is collectively represented as the color value COL _BL . In step 303, a comparison operation is performed, and the color COL _BL of the instantaneous block is compared with a predetermined color COL _H (skin color). Such a comparison operation is generally performed as follows.

| COL _H −COL _BL | <S (4) As described above, the color values are representative of lightness, hue, and saturation, and these are used as individual elements of the color value in the comparison operation. Used. Hereinafter, three different comparison operations will be described, but the present invention is not limited to these operations.

| Xy-hy | + | xCr-hCr | + | xCb-hCb | <S (1) | xy-hy | ² + | xCr-hCr | ² + | xCb-hCb | ² <S (2) k1 | D1 | + k2 | D2 | + k3 | D3 | <S (3) where xy is the luminance value (= lightness) of the first color value xCr the first chrominance value of the first color value (= hue) XCb Second chrominance value of first color value (= saturation) hy Luminance value of second color value (= brightness) hCr First chrominance value of second color value (= hue) hCb second The second chrominance value (= saturation degree) of the color value of S is the predetermined threshold value k1, k2, k3 The predetermined weight value D1 The first comparison between the luminance value of the first color value and the luminance value of the second color value D2 a second comparison of the first chrominance value of the first color value with a first chrominance value of the second color value D3 a second comparison of the second chrominance value of the first color value and the second chrominance value of the second color value 3 represents a third comparison with a chrominance value of 2.

In step 303, the result of the comparison operation is compared with a threshold value S. Predetermined color (skin color) C
If the difference between OL _H and the color COL _BL of the block is smaller than a predetermined threshold value S, the color COL _BL of the block is at least approximate to the skin color COL _H. In this case, a query is made in step 304 as to whether the instantaneous quantization value QW is equal to the first quantization value QW1. Note that the first quantization value QW1 is smaller than the second quantization value QW2. The lower the quantization value, the higher the image quality achieved by this. If it is determined in step 304 that the quantization value QW is equal to the first quantization value QW1, quantization is performed (see step 306). Otherwise, in step 305, the first quantization value QW1 is set as the quantization value QW, and subsequently, in step 306, quantization is performed.

If the similarity between the lock color COL _BL and the skin color COL _H cannot be obtained in the comparison operation 303, whether or not the instantaneous quantization value QW is equal to the second quantization value QW 2 is determined in step 307. Is asked. If they are equal, quantization is performed (see step 306). Otherwise, in step 308, the second quantization value QW2 is set as the new quantization value QW.

In steps 304 and 307 within the frame of the pre-processing described above, it is not only an inquiry as to whether or not the quantized value QW must be adapted as a result of the comparison operation from step 303, A large associated area (to be transmitted by quantization value) is obtained.

FIG. 4 shows a configuration for performing the quantization depending on the buffer state.

Block 401 of FIG. 4 is the first block diagram of a classic intraframe coder.
Is a group of blocks. The digital image is transformed into the frequency domain based on DCT, where spectral coefficients are weighted and quantized at block 402. After the quantization, in step 403, sampling by variable length coding is performed, and the data thus obtained is written to the buffer 404.
The compressed image data is transmitted from the buffer 404 to the decoder via the channel 406.

Although the data rate of the channel is constant, a predetermined time is used for each image to be transmitted, since image transmission must at least approximately meet real-time requirements. If no image encoding is performed at this time, the entire image will not be transmitted. To ensure that the entire image is always transmitted, the quantization 402 is adapted depending on the filling state of the buffer 404 (see adaptation path 405). If the image coding shows that the remaining time is not enough to transmit the entire image to the decoder, the quantization is adapted by increasing the quantization value QW. This results in inaccurate and unclear image compression, but can perform the compression at a correspondingly high speed and meet real-time requirements.

Here, the first region that meets the criteria and is to be transmitted with a correspondingly high quality is the remaining second region.
It is always guaranteed that the image quality is higher than that of the region. Therefore, the quantization value for the first area is always smaller than the quantization value for the second area.

FIG. 5 shows another means for improving image quality (see block 501).

In view of the method described in detail above, the image quality improvement is carried out, inter alia, on certain areas having a certain color. The skin tones of the image should be transmitted at a higher image quality than the rest of the image and presented to the decoder.

The first means 502 increases the image repetition rate (for regions having a predetermined color). By updating the corresponding area frequently, this area is displayed more smoothly than the rest of the motion picture. By a corresponding adjustment during the image coding, a real-time display of this region is advantageously ensured. As a result, the smooth movement of the flesh-colored surface is also perceived on the decoder side.

The second means 503 improves the position resolution. The image sharpness is improved by increasing the number of pixels per unit area for an area satisfying the predetermined criterion.

Finally, the third means 504 reduces the quantization value for a given area (shown in detail above). By reducing the quantization value compared to the rest of the image, the image coding for the region of interest, preferably of flesh color, is performed with fine quantization, and thus the image quality is improved.

The above-described image quality improving means for a predetermined area satisfying the criterion can be combined.

The following publications are cited within the context of the present description.

[0093]

[Table 1]

[Brief description of the drawings]

FIG. 1 shows a device having two calculators, a camera and a screen, by which the encoding, transmission, decoding and display of image data takes place.

FIG. 2 is a schematic diagram of an apparatus for encoding a digital image on a block basis.

FIG. 3 shows the steps of a method for quantizing a digital image by a calculator.

FIG. 4 is a schematic diagram showing how quantization is adapted depending on a buffer state.

FIG. 5 is a schematic diagram of a means for improving image quality.

[Procedural Amendment] Submission of translation of Article 34 Amendment of the Patent Cooperation Treaty

[Submission date] November 15, 1999 (November 15, 1999)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Claims

[Correction method] Change

[Correction contents]

[Claims]

──────────────────────────────────────────────────続 き Continued on the front page F-term (reference) CA27 DC06 DD06 GA02 GA05 HA01 HA03 JA03 KE16 KE17

Claims (41)

[Claims] 1. An image processing method comprising: a) dividing a digital image into two regions, a first region meeting a predetermined criterion, a second region not meeting a predetermined criterion, b) a first region An image processing method, wherein the area is processed with higher image quality than the second area. 2. The method according to claim 1, wherein the image is divided into a plurality of regions, and the first regions satisfying the criterion are processed with higher image quality than the second regions not satisfying the criterion. . 3. The method according to claim 1, wherein the criterion is that the area of the image has a color from a predetermined color area. 4. The method according to claim 3, wherein the color substantially corresponds to a color of a human skin. 5. The method according to claim 1, wherein the image is processed by a block-based image coding method. 6. A block-based image encoding method according to the MPEG standard or H.264 standard.
261 or H.264. The method according to claim 5, defined by the H.263 standard. 7.) a) detecting, for a block having a predetermined size of the image, the color of the block in the form of a second color value of the block; b) detecting the second color value and the first color value; Performing a comparison operation with a color value; c) the criterion is met if the result of the comparison operation is less than a predetermined threshold, ie the block is at least approximating flesh color; d) otherwise the criterion is the block 7. A method according to claim 5 or claim 6, wherein the method is not satisfied. 8. The method of claim 7, wherein the method is iteratively performed on each block of the image. 9. The method according to claim 7, wherein the predetermined size of the block is 8 pixels × 8 pixels. 10. The predetermined size of the block is 16 pixels × 16 pixels.
The method according to claim 7. 11. The comparison operation is performed by the following equation: | xy−hy | + | xCr−hCr | + | xCb−hCb | <S, where xy is the luminance value (= lightness) xCr of the first color value. The first chrominance value (= hue) of the first color value xCb The second chrominance value (= saturation) of the first color value hy The luminance value (= lightness) of the second color value hCr The second color 11. The method according to claim 7, wherein the first chrominance value of the value (= hue) hCb is the second chrominance value of the second color value (= saturation) S a predetermined threshold value. 12. The comparison operation is performed by the following equation: | xy−hy | ² + | xCr−hCr | ² + | xCb−hCb | ² <S where xy is the luminance value of the first color value (= Lightness) xCr First chrominance value of first color value (= hue) xCb Second chrominance value of first color value (= saturation) hy Luminance value of second color value (= lightness) hCr 11. The first chrominance value (= hue) of the second color value hCb The second chrominance value (= saturation) of the second color value S is a predetermined threshold value. the method of. 13. The comparison operation is performed according to the following equation: k1 | D1 | + k2 | D2 | + k3 | D3 | <S where k1, k2, k3 predetermined weighting values D1 first color value luminance value D2 a first comparison between the first chrominance value of the first color value and a first chrominance value of the second color value D3 the first color The method according to any of claims 7 to 10, wherein a third comparison S of the second chrominance value of the value and the second chrominance value of the second color value is a predetermined threshold value. 14. A block-based image coding method is defined according to the MPEG-4 standard, and an area defined according to the MPEG-4 standard satisfies said predetermined standard.
14. The method according to any one of the preceding claims. 15. The method according to claim 1, wherein the image quality is improved by increasing the image repetition rate for the first region. 16. The method according to claim 1, wherein the position resolution is increased for the first area, and the image quality is improved by processing a relatively large number of pixels for the first area.
The method according to any one of claims 1 to 15. 17. The image quality is improved by quantizing a first area with a first quantization value and quantizing a second area with a second quantization value. 17. The method according to any of the preceding claims, wherein is less than the second quantized value. 18. The switching from the first quantization value to the second quantization value is performed by DQ
By transmitting the UANT symbol as a parameter together with the second quantization value, the switching from the second quantization value to the first quantization value is performed by DQUANT.
The method according to claim 17, wherein the method is performed by transmitting the symbol as a parameter together with the first quantization value. 19. If the criterion is fulfilled, the following measures are carried out in order to guarantee as large a relevant area as possible with a plurality of blocks and with a predetermined color (skin color): a) The block does not fulfill the criterion If the result of the comparison operation on the block is slightly above a predetermined threshold value if the block is substantially surrounded by another block satisfying the criterion, the block is encoded with the first quantization value. And b) if the block does not satisfy the criterion and the block is located in the center of the block that satisfies the criterion, the block is also encoded with the first quantized value.
Or the method of 18. 20. An image processing apparatus having a processor unit, wherein: a) a digital image is divided into two regions, a first region satisfies a predetermined criterion,
An image processing apparatus, wherein the second area does not satisfy a predetermined criterion, and b) the first area is configured to be processed with higher image quality than the second area. 21. The processor unit, wherein the image is divided into a plurality of regions, and a plurality of first regions satisfying the criterion are processed with higher image quality than a plurality of second regions not satisfying the criterion. 21. The device of claim 20, wherein the device is configured. 22. The apparatus according to claim 20, wherein the processor unit is configured such that it is based on the fact that the area of the image has a color from a predetermined color area. 23. The apparatus of claim 22, wherein the processor unit is configured such that the color substantially corresponds to a color of a human skin. 24. The apparatus according to claim 20, wherein the processor unit is configured such that the image is processed by a block-based image coding method. 25. The processor unit according to claim 1, wherein the block-based image coding method is the MPEG standard or H.264. 261 or H.264. The apparatus of claim 24, wherein the apparatus is configured to be defined according to the H.263 standard. 26. A processor unit comprising: a) for a block having a predetermined size of an image, a color of the block is detected in the form of a second color value of the block; b) a second color value of the block; A) a comparison operation is performed with the first color value; c) if the result of the comparison operation is smaller than a predetermined threshold, the criterion is fulfilled, ie the block is at least approximating skin color; d) otherwise: 26. The apparatus according to claim 24 or claim 25, wherein the criterion is configured not to be met for the block. 27. The apparatus of claim 26, wherein the processor unit is configured to perform the method iteratively on each block of the image. 28. The apparatus according to claim 26, wherein the processor unit is configured such that the predetermined size of the block is 8 pixels × 8 pixels. 29. The apparatus according to claim 26, wherein the processor unit is configured such that the predetermined size of the block is 16 pixels × 16 pixels. 30. The processor unit, wherein the comparison operation is performed by the following equation: | xy−hy | + | xCr−hCr | + | xCb−hCb | <S where xy first Luminance value (= lightness) of the color value of xCr first chrominance value of the first color value (= hue) xCb second chrominance value of the first color value (= saturation) hy of the second color value The luminance value (= brightness) hCr the first chrominance value (= hue) of the second color value hCb the second chrominance value (= saturation) of the second color value S is a predetermined threshold value. 30. Apparatus according to any one of the preceding claims. 31. The processor unit is configured such that the comparison operation is performed according to the following equation: | xy−hy | ² + | xCr − hCr | ² + | xCb −hCb | ² <S xy luminance value of first color value (= brightness) xCr first chrominance value of first color value (= hue) xCb second chrominance value of first color value (= saturation) hy second The luminance value (= lightness) of the color value hCr The first chrominance value (= hue) of the second color value hCb The second chrominance value (= saturation) of the second color value S is a predetermined threshold value. Item 30. The apparatus according to any one of Items 26 to 29. 32. The processor unit is configured such that the comparison operation is performed according to the following equation: k1 | D1 | + k2 | D2 | + k3 | D3 | <S where k1, k2, k3 predetermined D1 a first comparison between the first color value luminance value and the second color value luminance value D2 a first chrominance value of the first color value and a first chrominance value of the second color value 30. A third comparison S3 with a second threshold value D3 of a second chrominance value of the first color value and a second chrominance value of the second color value. An apparatus according to any one of the preceding claims. 33. The processor unit, wherein the block-based image coding method is defined by the MPEG-4 standard, and a predetermined standard is satisfied for an area defined by the MPEG-4 standard. 33. Apparatus according to any one of claims 20 to 32. 34. The apparatus according to claim 20, wherein the processor unit is configured to improve image quality by increasing an image repetition rate for the first region. . 35. The processor unit is configured to increase positional resolution for the first region and to improve image quality by processing a relatively large number of pixels for the first region. 35. The device according to any one of claims 20 to 34, wherein 36. The processor unit, wherein the first area is quantized by a first quantization value and the second area is quantized by a second quantization value so that image quality is improved. 36. Apparatus according to any one of claims 20 to 35, wherein the first quantized value is configured to be less than the second quantized value. 37. The processor unit switches from the first quantized value to the second quantized value by transmitting a DQUANT symbol as a parameter together with the second quantized value, and vice versa. 37. The apparatus according to claim 36, wherein switching from the quantized value to the first quantized value is performed by transmitting a DQUANT symbol as a parameter together with the first quantized value. 38. The processor unit comprises a plurality of blocks when the criterion is fulfilled, and is configured such that the following measures are performed in order to guarantee as large an associated area having a predetermined color (skin color). A) if the block does not meet the criterion and if the block is almost surrounded by another block that satisfies the criterion, the result of the comparison operation on the block is slightly above a predetermined threshold value; B) encode the block with the first quantized value; b) if the block does not meet the criterion and the block is located in the center of the block that satisfies the criterion, the block is also encoded with the first quantized value Claim 36
Or the method of 37. 39. A means is provided for: a) performing a spectral transformation, said means for converting a digital image into the spectral domain; b) providing means for entropy encoding, said means comprising: Performing compression, c) a buffer is provided, said buffer receiving data of variable data rate from the means for entry-pey encoding and further conducting to a channel of fixed data rate. The device according to any one of claims 20 to 38. 40. The apparatus according to claim 39, wherein the processor unit is adapted to adapt the quantization value depending on a filling state of the buffer. 41. The processor unit, wherein the quantization value is adapted such that the quantization value is increased when the buffer is full, thereby blurring the image transmission. 41. Apparatus according to claim 40, wherein the quantization value is reduced when the buffer is empty, so that a relatively high image quality is ensured.